Google unveils Gemini 3.5, Spark

Google used its I/O conference on May 19 to widen the gap between chatbots that answer questions and software that carries out tasks. The company introduced Gemini 3.5, starting with a 3.5 Flash model, and a new agent called Gemini Spark that it said can manage work across connected apps and services. Sundar Pichai announced the products in Mountain View, California, as Google pressed its case that AI should move from prompts to actions. Company blog posts and CNBC reports showed Google tying the rollout to the Gemini app, AI Mode in Search, developer tools and Chrome. ### So what is Gemini 3.5, exactly? Google said Gemini 3.5 is a new family of models “combining frontier intelligence with action,” with 3.5 Flash as the first release. The company said 3.5 Flash is available immediately in the Gemini app, AI Mode in Google Search, Google AI Studio, Android Studio and enterprise products, while Gemini 3.5 Pro is being used internally and is due next month. (blog.google) Koray Kavukcuoglu, Jeff Dean, Oriol Vinyals and Noam Shazeer wrote in Google’s launch post that 3.5 Flash is built for “complex, agentic workflows.” Google said the model outperforms Gemini 3.1 Pro on several coding and agent benchmarks and runs four times faster than other frontier models, though those comparisons came from Google’s own cited benchmarks. ### What does Gemini Spark do that a chatbot does not? (blog.google) Josh Woodward, Google Labs’ vice president for the Gemini app and AI Studio, described Gemini Spark as “a 24/7 personal AI agent” built to proactively manage tasks. Google said Spark is meant to operate under user direction and help people navigate their digital lives, rather than only respond to one-off prompts. (blog.google) CNBC reported Google presented Spark as a general-purpose agent in the Gemini app that can reason across information in connected apps. In Google’s account of the launch, Spark sits alongside other new Gemini app features including Daily Brief, a redesigned interface and deeper voice and desktop functions. ### Where do third-party services fit in? Investing.com, cited in the source briefing, reported Gemini Spark will use MCP, or Model Context Protocol, to connect to outside services including DoorDash, Uber, Lyft, Instacart, Booking Holdings and Expedia partners. (blog.google) Google’s official launch materials available in search results did not list that full partner set in the excerpts I could verify directly, so that detail remains reported by secondary outlets rather than confirmed in the Google pages reviewed here. Google has already been moving in that direction elsewhere. In a February product post, the company showed Gemini handling tasks such as booking a ride home or reordering a DoorDash meal on Android devices, and Chrome materials released around I/O described “Connected Apps” across Gmail, Calendar, YouTube, Maps, Shopping and Flights. ### Why is Chrome part of this rollout? (cnbc.com) Parisa Tabriz, Google’s vice president for Chrome, said the browser is getting new Gemini features including a side panel, deeper app integrations and “auto browse.” Google said those updates are built on Gemini 3 and are designed to help users compare tabs, summarize information and carry out multi-step workflows on the web. Google’s Chrome post said more contextual help through “Personal Intelligence” is coming in the months ahead. (blog.google) That places the browser as one of the main surfaces where Google is trying to turn Gemini from an assistant into an agent that can work across tabs, apps and web tasks. ### How big is Google saying Gemini has become? Google said Gemini now reaches more than 900 million monthly users across 230 countries and more than 70 languages. (blog.google) Pichai also said Google is processing more than 3.2 quadrillion tokens per month across its surfaces and that more than 8.5 million developers are building with its models each month. Those figures were part of Google’s effort to show scale as OpenAI and Anthropic compete for developers, enterprise spending and consumer attention. (blog.google) CNBC reported Wall Street has been looking for deeper integrations across Google products after the company’s rising capital spending, and Google is presenting agents as one answer. ### When can people actually use these features? Google said Gemini Spark will be available first to trusted testers and Google AI Ultra subscribers starting next week. (blog.google) CNBC reported the same timing for the beta rollout, while the source briefing said Chrome availability is expected later this summer. Google said Gemini 3.5 Flash is already live across the Gemini app, AI Mode in Search and developer products, and said Gemini 3.5 Pro is scheduled for wider rollout next month. (cnbc.com) Chrome’s newer Gemini workflow features are due in the coming months, according to Google’s browser update. (blog.google)